Wrong class inheritance with self.__class__? [duplicate]

Question

This question already has an answer here:

How to avoid infinite recursion with super()? (1 answer)

Closed 3 years ago.

I was working with inheritance in WTForms. I had next class:

class MyForm(WTForms):
    ...
    def process(self, formdata=formdata, obj=None):
        super(self.__class__, self).process(formdata=formdata, obj=obj)

And I had error (not always, but in some cases with similar forms - it is even more strange):

Maximum recursion depth have been exceeded

But when I've changed self.__class__ → MyForm everything was OK!

I don't get it... As I know they must be synonyms. What's wrong?

Answer 1

They are not synonyms. self may be any object of a class that inherits from MyForm. Python does not silently create a new superclass object for calling inherited methods, it just passes the same object (doing so would break polymorphism and serve no use). And the type()/.__class__ is obviously the class the object is really an instance of, not some superclass of that (you don't expect MyShinyThing().__class__ to yield object, right?). It would be really astonishing, non-idiomatic and useless if __class__ changed depending on where it is accessed from. It's just polymorphism.

And as the first argument to super is an indicator where in the MRO the search for a supermethod should continue (informally: where you currently are), passing self.__class__ always starts back at the (immediate) superclass of whatever class self is an instance of. Assuming proper super calls elsewhere in the class hierarchy, this ultimately gets you back to MyForm.process, and we've got our infinite loop.